karpathy-graph-query skillA
karpathy-graph-query is agent-read markdown (skill) from andersonlimahw/lemon-ai-hub: Stage 4 of the karpathy-graph pipeline. Answer multi-hop questions from a knowledge graph by retrieving a bounded subgraph with edge citations, and ground or refute claims against actual graph paths. Use when the user asks a question that spans several facts, asks how two things connect, asks to verify or ground a claim, or when an agent needs task-specific context from an existing graph..
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# Query
The graph exists so agents stop replaying transcripts. That only works if
retrieval stays **bounded**.
## The rule
**Never dump the graph into a prompt.**
Not `graph.json`, not the full node list, not "just this once because it's
small". Retrieve the connected state needed for the current decision, and
nothing else. A graph used as a context dump is worse than no graph — it costs
the same tokens plus the build.
## Retrieval procedure
1. Resolve the entities the question mentions
2. Expand one or two hops over the edge types that matter
3. Filter by type, source, or date when the question implies it
4. Serialize within a token budget
5. Keep the edge ids so the answer can cite them
```bash
KG="${CLAUDE_PLUGIN_ROOT:-plugins/karpathy-graph}/scripts/kgraph.py"
python3 "$KG" --name "<graph>" query "vendor X" \
--hops 2 \
--edge-type SUPPORTS --edge-type MENTIONS \
--max-nodes 40 \
--format context
```
## Choosing hops
| Hops | Use for |
|---|---|
| 1 | "what is directly attached to X" |
| 2 | most questions — the default |
| 3 | explicit multi-hop chains (X → Y → Z) |
| 4+ | almost never; you are retrieving the whole graph with extra steps |
…Read the whole file at its exact version.
How to install
mdr add andersonlimahw/lemon-ai-hub/karpathy-graph-query@v1.0.0mdr add andersonlimahw/lemon-ai-hub/karpathy-graph-query@sha256:a06c8fea87ea42fcPin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_b3gl4gg72yz3lbdt)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (4369 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
andersonlimahw/lemon-ai-hub · 14 stars · license none · pushed 2026-09-20 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_b3gl4gg72yz3lbdt GET https://markdownregistry.com/api/v1/resolve?ref=andersonlimahw/lemon-ai-hub/karpathy-graph-query GET https://markdownregistry.com/api/v1/blob/a06c8fea87ea42fcb64cf4fea1902ae908db4c2527f5825798b6bedd7b645af7
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.